International Trends and Influencing Factors in the Integration of Artificial Intelligence in Education with the Application of Qualitative Methods

Abstract

This study offers a comprehensive examination of the scientific output related to the integration of Artificial Intelligence (AI) in education using qualitative research methods, which is an emerging intersection that reflects growing interest in understanding the pedagogical, ethical, and methodological implications of AI in educational contexts. Grounded in a theoretical framework that emphasizes the potential of AI to support personalized learning, augment instructional design, and facilitate data-driven decision-making, this study conducts a Systematic Literature Review and bibliometric analysis of 630 publications indexed in Scopus between 2014 and 2024. The results show a significant increase in scholarly output, particularly since 2020, with notable contributions from authors and institutions in the United States, China, and the United Kingdom. High-impact research is found in top-tier journals, and dominant themes include health education, higher education, and the use of AI for feedback and assessment. The findings also highlight the role of semi-structured interviews, thematic analysis, and interdisciplinary approaches in capturing the nuanced impacts of AI integration. The study concludes that qualitative methods remain essential for critically evaluating AI’s role in education, reinforcing the need for ethically sound, human-centered, and context-sensitive applications of AI technologies in diverse learning environments.

1. Introduction

In recent years, Artificial Intelligence (AI) has rapidly emerged within the field of education, transforming not only teaching and learning processes but also research methodologies [1]. Specifically, qualitative data analysis, traditionally associated with interpretive and craft-based approaches, has begun to benefit from AI-powered tools that offer increased efficiency, objectivity, and processing capacity [2]. These innovations allow for the automation of tasks such as large-scale data coding, thematic pattern recognition, and sentiment analysis, thereby opening new possibilities while also raising ethical and epistemological challenges [3].

Within this context, the integration of AI into qualitative educational research has become an emerging international trend. However, questions remain regarding the factors that influence its adoption and the impact of these technologies on the validity and depth of qualitative analysis [4]. Understanding how and why these tools are implemented, along with their methodological and ethical implications, is essential for the development of rigorous and responsible research practices in education.

The aim of this study was to explore international trends and key factors influencing the integration of AI into the field of education through qualitative research methods, considering both the potential and the challenges this integration poses for contemporary educational inquiry. To achieve this objective, this study conducts a bibliometric analysis, a methodology particularly suited to identifying emerging trends, influential authors, and collaboration networks, of research on AI-assisted qualitative analysis. This work seeks to bridge existing knowledge gaps regarding the evolution of academic perspectives and practical applications in AI-mediated qualitative research. The study addresses the following research questions:

• How has scientific research output on the integration of AI in education using qualitative methods evolved over the past decade?
• Which journals, authors, and sources are the most influential in research on AI in education linked to qualitative methods?
• Which countries and institutions lead in productivity and collaboration in the study of AI applications in education through qualitative data analysis?
• What are the main thematic trends that have characterized qualitative research on the integration of AI in education in recent years?

2. Theorical Framework

2.1. Foundations and Early Developments in the Use of AI in Qualitative Educational Research (2014–2019)

The integration of AI into qualitative educational research began to gain traction in the mid-2010s, as advances in natural language processing (NLP), text mining, and machine learning opened new possibilities for analyzing unstructured data. Between 2014 and 2019, a number of pioneering works explored the potential of AI tools to support, complement, or even partially automate stages of the qualitative research process. Early studies focused primarily on two domains: (1) the use of NLP for coding and categorization of open-text responses [5,6] and (2) the application of machine learning algorithms to detect themes and patterns in large-scale qualitative datasets [7,8]. These initial efforts were particularly relevant in educational contexts involving open-ended survey data, student feedback, or digital ethnographies.

Simultaneously, methodological discussions emerged regarding the implications of algorithmic support for researcher reflexivity and interpretative depth [9,10]. These debates laid the foundation for the later emergence of critical perspectives on the role of AI in the epistemology of qualitative inquiry. Although the full integration of AI into qualitative research design was still incipient, this period laid essential groundwork for the post-2020 expansion, as technological accessibility improved and educational data grew in complexity and volume [11].

2.2. AI in Education: Advances, Emerging Applications, and Challenges

AI has revolutionized education [12] by becoming integrated into multiple key areas (Figure 1), enhancing both teaching practices and academic management systems. One of its most impactful applications lies in curriculum design, where advanced algorithms analyze educational trends and generate personalized learning materials tailored to diverse learning styles [13]. In the domain of personalized learning, AI enables the development of adaptive platforms that adjust content difficulty and pacing based on individual student needs, incorporating chatbots and virtual assistants for automated tutoring and instant feedback [14]. Specifically, according to [15], there are three approaches that can be integrated into a personalized, level-based learning system: (1) manual leveling, where teachers configure differentiated learning paths based on initial diagnoses; (2) scheduled leveling, based on predefined rules that adjust user progression based on their performance; and (3) dynamic adaptive systems, which incorporate AI algorithms to adjust content, activities, and sequences in real time according to the student profile and responses. In this way, the practical impact of the proposed system is strengthened, aligning it with current trends in user-centered educational software development.

Furthermore, AI-powered assessment tools can automatically grade assignments and exams, offering detailed reports on student progress and areas requiring improvement [16,17]. These applications, as shown in the review in [18], are particularly relevant in marking open-ended responses and essays, where traditional systems, such as those integrated into platforms like Moodle, have limitations by operating only with closed-ended items or giving feedback based on predefined parameters. Unlike these simpler systems, AI can tackle more complex tasks using natural language processing (NLP) techniques and deep neural network models, which analyze not only the grammatical structure of responses but also their coherence, argumentation, style, and semantic content.

Among the most advanced solutions are e-rater (ETS), used in standardized tests, such as the TOEFL and WriteToLearn, which combines latent semantic analysis with statistical models to assess written essays. In Spanish-speaking contexts, tools such as WriteWise have begun to be integrated into university environments for similar purposes. These systems not only grade responses but also provide automated formative feedback, allowing students to review and improve their textual production before a final assessment.

From a development perspective, these applications rely on widely used NLP environments and libraries, such as spaCy, NLTK, BERT, GPT, and the pre-trained models available in Hugging Face Transformers, which offer robust capabilities for understanding context, detecting discursive errors, and generating automatic assessments aligned with teacher rubrics. The inclusion of these technologies represents a step toward more personalized and scalable assessment, although it must always be complemented by teacher supervision and adaptation to the educational context, avoiding excessive reliance on algorithms without adequate ethical and pedagogical validation [19,20].

In the field of predictive analytics and educational diagnostics, machine learning models are used to identify academic performance patterns and detect at-risk students early on, facilitating timely and personalized interventions [21]. At the institutional level, academic and administrative management is enhanced through automation of tasks such as admissions, resource allocation, and student data management, significantly reducing bureaucratic workload [22]. As noted in reference [23], applying AI to curriculum design allows us to move beyond uniform approaches to educational planning by incorporating elements of learning analytics, student profile clustering, and predictive modeling. These tools can help analyze the historical performance of previous cohorts, identify patterns of success or difficulty, and suggest differentiated curricular paths. For example, clustering algorithms allow students to be grouped based on cognitive, motivational, and contextual variables, facilitating the design of personalized learning paths. Furthermore, models such as recurrent neural networks (RNNs) or random forests can predict, with a high degree of accuracy, the probability of dropout, the optimal cognitive load, or the most effective sequence of content. These predictions are used to reorganize the curricular sequence or recommend specific teaching resources [24].

The use of libraries such as TensorFlow, Keras, and Scikit-learn allows for the development of systems that integrate these models into learning management systems (LMSs). Some pilot projects at European universities are already using these approaches to redesign courses in real time, generating personalized dashboards for both students and instructors. Although adaptive curriculum design may seem like an extension of personalized learning, its focus is more structural: it seeks to reconfigure the content itself based on profiles and predictions, not just adjust the delivery interface [25]. Therefore, it represents a complementary and strategic contribution within the AI-powered educational ecosystem.

Regarding accessibility and educational inclusion, AI supports content adaptation for students with disabilities through speech recognition, automated translation, and text-to-speech technologies, promoting more equitable learning opportunities [26]. Additionally, data-driven learning and gamification increase student motivation and engagement by leveraging big data analytics to inform pedagogical decision-making and design immersive, interactive experiences [27]. Finally, AI also contributes to teacher training and professional development by offering self-assessment tools, data-informed pedagogical strategy recommendations, and personalized training pathways [28]. Collectively, these applications position AI as a transformative technology capable of making education more efficient, inclusive, and responsive to the evolving needs of the 21st century.

However, the integration of AI in education presents a range of challenges and limitations that must be addressed to ensure its effective and equitable implementation. One of the foremost concerns relates to ethical and equity issues, as AI algorithms can perpetuate bias if the training data are not representative of diverse populations [29]. Additionally, unequal access to AI technologies risks widening the educational gap, disproportionately affecting students and schools with fewer resources [30]. The AI tools that generate the greatest inequality in education and society are those that intensify the digital divide, reproduce algorithmic biases, and present access barriers for vulnerable populations. In particular, AI-driven automation systems contribute to the labor displacement of lower-skilled individuals, reducing their ability to acquire new digital skills and access emerging technologies [31]. Furthermore, the algorithmic biases present in educational or assessment platforms, such as biases in content recommendation systems, automated assessment with linguistic or cultural biases, biases in dropout risk prediction, or algorithmic discrimination in intelligent tutors, reinforce historical inequalities by indirectly discriminating against marginalized groups, as has been documented in different social contexts [32]. Generative tools, such as ChatGPT, also widen the gap between those with digital access and those without, as their use is concentrated in urban, advanced socioeconomic sectors, while their adoption is much lower in rural or disadvantaged regions [33].

Another critical challenge is data privacy and security [34]. Implementing AI in educational settings requires collecting and analyzing large volumes of personal information from students and educators, posing risks in terms of data protection. Compliance with privacy regulations, such as the GDPR in Europe and FERPA in the United States, is essential to prevent the misuse or leakage of sensitive information [35]. Technological dependency and the digital divide also hinder effective adoption. Not all educational institutions possess the necessary infrastructure to implement AI systems effectively [36], and both teachers and students may lack the training required to use these tools optimally [37], resulting in unequal learning opportunities and diminished outcomes [38].

The evolving role of teachers is another frequently cited concern. The automation of educational tasks may lead to uncertainty about educators’ place in the classroom, especially if AI is perceived as a replacement for human interaction in teaching [22]. It is therefore vital that educators receive targeted professional development to integrate AI into their pedagogical practices without losing their role as learning facilitators. The current lack of regulations and standards for AI use in education further complicates its safe and effective deployment [39]. Without universal frameworks guiding the design, implementation, and evaluation of AI in educational environments, applications may be inconsistent and exacerbate inequalities [40].

Another significant issue involves the interpretability and transparency of AI models. Many systems function as “black boxes,” meaning users may not understand how decisions or recommendations are made. This lack of algorithmic explainability can foster distrust and hinder broader acceptance in educational contexts [41]. Cost and sustainability also represent substantial barriers. Effective AI integration requires major investments in hardware, software, and teacher training, which are resources that not all institutions can afford, raising questions about long-term viability [42]. The distinction between AI systems that require significant investment and those that are simpler and more accessible in education can be illustrated through specific examples. High-investment systems include learning analytics platforms based on big data, intelligent tutoring systems, and adaptive learning environments. These tools demand robust infrastructure, powerful servers, ongoing maintenance, and specialized personnel to manage and interpret the data effectively [43]. Such technologies are typically suited for well-resourced institutions, such as large universities or centralized education systems. In contrast, more accessible AI tools include generative AI applications for writing support, virtual assistants based on natural language processing, and mobile learning platforms with basic personalization features. Many of these are available through open-source formats or free educational licenses, making them feasible for institutions with limited budgets and infrastructure [44]. This distinction allows educational institutions to align AI integration with their actual technological and financial capacities, promoting a more equitable and sustainable use of AI in education.

Finally, resistance to change remains a limiting factor in AI adoption. Teachers, administrators, and students may hesitate to embrace AI tools due to lack of awareness or fears that technology may displace traditional teaching methods [45]. Attitudes toward the adoption of AI tools in education show substantial differences between students and teachers, influenced by cultural, institutional, technological, and generational factors. While students often perceive AI as a useful tool to personalize and enhance their learning experience, educators tend to assess these technologies more critically, focusing on their ethical, pedagogical, and professional implications [46]. In general, students—especially in higher education contexts—value AI for its ability to provide immediate feedback, facilitate access to content, and support self-directed learning. Many see AI as a means to foster autonomy and analyze efficiency, in addition to using it for problem-solving beyond the classroom [47]. This positive outlook is often associated with greater familiarity with digital tools and a more pragmatic attitude toward technology, particularly among younger generations [48].

In contrast, educators express more complex concerns. On one hand, there are worries about AI’s impact on academic integrity, specifically predicated on the risk of increased plagiarism and the misuse of generative tools [49]. On the other hand, some teachers fear a potential dehumanization of the educational process, the loss of pedagogical control, or students becoming overly dependent on technology [50]. These concerns are more common among experienced teachers, who often show greater resistance to adopting new technologies and tend to prioritize traditional teaching methods. Younger educators or those with stronger digital competencies, however, tend to be more open to integrating AI into their teaching practices [51]. Cultural context also plays a key role. Significant differences in trust and willingness to use AI can be observed across countries and regions. In environments where educational innovation is a political priority and technological investment is strong, both teachers and students demonstrate more favorable attitudes [52]. In other cases, poor infrastructure, limited technological training, or fear of teacher replacement serve as barriers to adoption.

To ensure successful integration, it is necessary to promote a culture of educational innovation and provide training that helps stakeholders understand both the benefits and limitations of AI. These challenges underscore the need for strategic, regulated, and ethical AI implementation in education, ensuring that its advantages outweigh the associated risks and limitations.

2.3. Integrating AI into Qualitative Data Analysis: Current Methodological and Ethical Perspectives

The emergence of AI in the field of qualitative research is shaping new methodological dynamics that promise to significantly transform traditional analytical practices. The integration of AI tools presents both disruptive possibilities and substantial ethical challenges. From a methodological standpoint, AI broadens the scope for researchers by automating routine tasks, assisting in the processing of large volumes of data, and facilitating the triangulation of qualitative information. One of the most evident examples of AI integration in qualitative research is its application for the automatic transcription of interviews, as demonstrated in [53]. The comparison of various AI platforms for transcribing interviews in Spanish highlights not only increased efficiency and transcription quality but also the importance of institutional context and accessibility when selecting tools. Beyond automation, these platforms allow researchers to focus more time on complex interpretative tasks, though the necessity of human review to ensure transcription fidelity remains essential.

Another area of application is the cleaning of qualitative databases, as illustrated in [54]. Through the use of AI tools, even novice researchers can transform unstructured data into analyzable formats, streamlining data preparation and fostering pedagogical opportunities for developing computational thinking and digital competencies in qualitative contexts. In terms of analysis and data extraction, studies such as [55,56] show that AI can identify trends, frequencies, and semantic patterns that largely align with human coding results. While some discrepancies are reported, these differences can enrich methodological debates and open new paths for triangulation and validation, as argued in [57]. In this way, AI is positioned not as a replacement but as a complement to human interpretive judgment, promoting more robust and transparent analyses. Several AI applications have demonstrated a strong capacity to identify trends, frequencies, and semantic patterns in coding data, often yielding results comparable to those produced by human coders. These tools harness advances in NLP and machine learning to interpret code not just as technical syntax but as a form of human expression, revealing language-like patterns [58]. For instance, pattern mining techniques have evolved to detect high-utility and statistically significant patterns that are both meaningful and human-interpretable [59]. Deep learning models, such as BiLSTM, have proven particularly effective in educational contexts, achieving classification accuracies of up to 95.97% when retrieving relevant code snippets [60]. Additionally, explainable AI approaches like EX-CODE enhance interpretability by using token occurrence probabilities to distinguish between human- and AI-generated code, offering transparent reasoning behind their decisions [61]. Despite these advancements, the discussion continues around AI’s limitations in capturing the deeper layers of human creativity and intuition involved in qualitative coding.

Regarding the validation of qualitative analysis, AI emerges as a promising tool for enhancing reliability and methodological transparency. References [62,63] suggest that AI can assess the consistency of coding processes, reduce individual bias, and improve efficiency without compromising rigor. Hybrid models, which foster collaboration between human coders and generative AI agents, are particularly promising. They not only increase efficiency but also deepen the understanding of studied phenomena by offering novel interpretative frameworks and enabling categorical redefinitions.

Nevertheless, these technological advances also bring ethical concerns that must be rigorously addressed. The recent literature emphasizes the need for transparency in the use of AI tools, as well as explicit acknowledgment of their role in analytical processes [64]. Risks include the reproduction of biases, data fabrication or “hallucinations” by generative models, and a tendency to conform to user expectations, which may distort findings. The equitable representation of marginalized or underrepresented voices is also a major concern, especially if AI-generated outputs are uncritically accepted as objective truths [65].

The most commonly used AI tools for qualitative data analysis include NVivo, ChatGPT, EPPI-Reviewer, Abstrackr, and AQUA, each offering specific functionalities that have significantly transformed traditional qualitative research methods. NVivo stands out for its ability to organize, code, and visualize data, making it widely used in complex projects [66]. ChatGPT, as a generative tool, has gained popularity for assisting in text-based data analysis, literature screening, and idea generation, although its outputs can be inconsistent and require careful oversight [67]. EPPI-Reviewer and Abstrackr are specifically designed for evidence reviews, enhancing productivity in the synthesis of qualitative data [68]. AQUA functions as an automated qualitative assistant, employing advanced coding techniques that significantly reduce processing time [69]. However, despite these advantages, researchers must use these tools with caution, as unsupervised application may compromise the integrity of analysis and overlook nuanced insights that only human judgment can fully capture.

3. Materials and Methods

3.1. Research Design

The methodological framework of this study was based on a Systematic Literature Review (SLR), complemented by bibliometric and content analysis techniques [70,71]. The integration of these approaches enabled a rigorous and comprehensive evaluation of the relevant scientific research output, as well as a detailed examination of its content. This methodological combination facilitated the identification of emerging trends, influential authors, and dominant lines of inquiry within the field of study [72].

3.2. Search Strategy and Document Selection

The search and selection process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [73], which offer a structured and rigorous framework for conducting systematic reviews. These guidelines emphasize transparency, replicability, and scientific rigor, ensuring that the review process is complete, traceable, and reproducible by other researchers. The workflow is visually presented in Figure 2, outlining the systematic stages of the review. Adherence to PRISMA guarantees methodological soundness and alignment with academic best practices [74].

This review focused on studies published in the last decade exploring the integration of AI into qualitative research in educational contexts, analyzing such studies using bibliometric approaches and visualization techniques. The bibliometric analysis enabled the identification and evaluation of scientific output on the topic of AI use in research based on publication patterns and impact metrics. To ensure the quality and relevance of the studies, documents indexed in the Scopus database were selected according to the following exclusion criteria:

• Not published between 1 January 2014, and 31 December 2024;
• Not written in English;
• Not published in a peer-reviewed academic journal.

A systematic search was conducted in Scopus using key terms such as “artificial intelligence,” “qualitative methods,” “qualitative analysis,” and “education,” applied to the fields of title, abstract, and keywords using the TOPIC field. The final search query used was as follows: (“artificial intelligence” OR “machine learning” OR “deep learning”) AND (“qualitative analysis” OR “qualitative data” OR “qualitative research” OR “text analysis”) AND (“education” OR “educational research”). Scopus was chosen due to its advanced visualization and analysis tools, making it an ideal source for identifying research trends in social sciences and education.

During the identification phase, 1120 records were retrieved. Ten records were removed prior to screening due to duplication or incomplete information. In the screening phase, 1110 documents were assessed, with 148 being excluded for falling outside the defined date range (1 December 2014–31 December 2024), 34 being excluded for not being written in English, and 298 being excluded for being published in non-peer-reviewed sources. As a result, 630 articles were deemed eligible and included in the final systematic review. This process ensured the selection of relevant, high-quality studies for analyzing the application of AI in education through qualitative methods.

The resulting database (Figure 3) reveals a significant increase in scientific research output, with an annual growth rate of 51.71%. A total of 630 documents were collected from 402 distinct sources, indicating a wide diversity of academic dissemination channels. The participation of 2351 authors reflects the high degree of collaboration in the field, with an average of 3.95 co-authors per document. Notably, only 87 publications were authored by a single researcher, suggesting that research in this area is predominantly conducted in collaborative teams.

Overall, 22.54% of the studies were the product of international collaboration, highlighting a growing trend toward cross-border research partnerships. In terms of impact, the documents have accumulated an average of 14.7 citations per publication, underscoring their influence within the scientific community. The analysis is further supported by a total of 29,837 references, indicating a robust and comprehensive bibliographic foundation. The average time since publication of 2.33 years suggests that research on AI in education is a relatively nascent and rapidly evolving field. This paper’s subsequent sections will examine the patterns of collaboration and thematic evolution within the literature, as well as the main contributions in the literature.

3.3. Data Analysis

The data analysis in this literature review was conducted in two main stages [75]. First, a descriptive analysis was performed to characterize the temporal evolution of scientific research output in this area. This involved identifying the number of publications per year, the distribution of authors and their contributions, and the most influential journals and high-impact documents. This approach provided a comprehensive overview of the field’s dynamics and growth, allowing for the identification of key trends and research gaps.

In the second stage, bibliometric visualization tools such as VOSviewer version 1.6.20 were used to construct knowledge networks. The first analysis focused on international collaboration through the development of co-authorship networks, where each node represented a country, and the links illustrated the frequency and strength of their collaborative efforts. Subsequently, a keyword co-occurrence analysis was conducted based on terms extracted from the selected documents. According to [76], this type of analysis facilitates an understanding of the conceptual and thematic structure of the research area, as well as the relationships between key terms. For this study, VOSviewer was used to generate the networks, applying a minimum co-occurrence threshold of 15 keywords and a relevance score of ≥ 60%. In the resulting network, each node represented a keyword, with node size reflecting frequency of appearance and link thickness indicating the strength of co-occurrence between terms.

Additionally, the analysis was complemented by the generation of thematic maps using the Bibliometrix package in RStudio version 2024.09.0+375. As noted by [77], these maps help visualize the distribution and relevance of topics within the research field through a two-dimensional diagram that groups keywords according to their density and centrality. Thematic clusters were identified, including motor themes as well as foundational and transversal themes, which are considered essential for the consolidation of the research area due to their high centrality and density, indicating their significance and strong interconnection with other core concepts in the field.

4. Results

4.1. Evolution of Scientific Research Output on the Integration of AI in Education from a Qualitative Approach (RQ1)

Over the past decade, AI has experienced exponential growth across various fields of knowledge, including education. In particular, the use of qualitative methodologies to analyze AI applications in educational settings has gained significant relevance (Figure 4). The findings reveal a progressive upward trend in the number of publications, beginning with a slow developmental phase between 2014 and 2019, during which annual output remained relatively low (ranging from 4–19 documents). Starting in 2020, there is a noticeable acceleration in scholarly output, with 27 publications that year, followed by 41 in 2021 and 68 in 2022. The most pronounced increases occurred in 2023 and 2024, reaching 117 and 323 documents, respectively, clearly reflecting an exponential growth trajectory.

4.2. Authors, Journals, and Reference Documents in Qualitative Research on AI in Education (RQ2)

Among the most influential scholars in the field (

Table 1. List of the 10 most productive authors in the field of AI in education using qualitative methods, emphasizing their contributions and influence based on the number of publications, h-index, and total citations (SPs = searched publications; TPs = total publications; and TCs = total citations).

	Author	SPs	TPs	h-Index	TCs	Current Affiliation	Country
1	Magana, Alejandra J.	4	214	26	2161	Purdue University	United States
2	Bannister, Peter	3	7	7	36	International University of La Rioja	Spain
3	Kantor, Jonathan	3	167	23	2833	University of Oxford	United States
4	Nanda, Gaurav	3	39	8	256	Purdue University	United States
5	Abisado, Mideth B.	2	100	9	268	National University	Philippines
6	Zou, Di	2	193	35	5254	Lingnan University	Hong Kong
7	Zheng, Lanqin	2	78	20	1213	Beijing Normal University	China
8	Zhai, Xiaoming	2	71	21	1632	University of Georgia	United States
9	Wulff, Peter	2	18	12	255	Pädagogische Hochschule Heidelberg	Germany
10	Williamson, Victoria	2	79	21	1896	King’s College London	United Kingdom

), Alejandra J. Magana from Purdue University (United States) stands out, leading the list with 4 publications specifically focused on the area, a total of 214 publications overall, an h-index of 26, and 2161 citations. Her academic trajectory encompasses a well-established contribution to research on the intersection of AI and education. Another prominent researcher is Jonathan Kantor from the University of Oxford (United Kingdom), with 167 publications, an h-index of 23, and 2833 citations, demonstrating a significant impact within the academic community. Equally noteworthy is Di Zou from Lingnan University (Hong Kong), who holds the highest number of citations (5254) and the highest h-index (35), indicating substantial influence in the fields of AI and education. Scholars such as Lanqin Zheng (Beijing Normal University, China) and Xiaoming Zhai (University of Georgia, United States) also exhibit high citation counts and elevated h-index scores, underscoring the consolidation of their research in this domain.

From a geographic and institutional perspective, the data reflect a strong presence of researchers from the United States, Europe, and Asia, suggesting an international distribution of knowledge in this field. Institutions such as Purdue University, the University of Oxford, Lingnan University, and Beijing Normal University emerge as key hubs in the production of research on AI in education using qualitative methods. Although some researchers may have fewer publications specifically in this area, their broader influence on AI and education remains significant, as evidenced by high citation metrics and scholarly impact. Furthermore, the involvement of authors from China, Hong Kong, and the Philippines highlights Asia’s increasing importance in terms of contributing to research on AI in education.

Our analysis of the most productive journals (

Table 2. The top 12 highly productive journals on research on AI in education using qualitative methods in the years (2014–2024) (SPs = searched publications; CDs = citations to date (2024); and BQ = best SJR (2023 quartile)).

	Journal	SPs	CDs	CiteScore 2023	SJR 2023	BQ	Publisher
1	PLOS One	13	337.945	6.2	0.839	Q1	Public Library of Science
2	Computers and Education: Artificial Intelligence	11	8.810	18.8	3.227	Q1	Elsevier
3	Journal of Medical Internet Research	11	45.742	14.4	2.020	Q1	JMIR Publications Inc.
4	Education and Information Technologies	10	28.189	10	1.301	Q1	Springer Nature
5	BMC Medical Education	9	17.254	4.9	0.935	Q1	Springer Nature
6	International Journal of Environmental Research and Public Health	8	328.899	7.3	0.808	Q2	MDPI
7	Nurse Education in Practice	7	5.352	5.4	0.869	Q1	Elsevier
8	Education Sciences	6	23.307	4.8	0.669	Q2	MDPI
9	Frontiers in Education	6	13.136	2.9	0.627	Q2	Frontiers Media S.A.
10	Frontiers in Psychology	6	135.179	5.3	0.800	Q2	Frontiers Media S.A.
11	International Journal of Learning, Teaching and Educational Research	6	2.564	2.1	0.287	Q3	Society for Research and Knowledge Management
12	BMJ Open	6	78.823	3.9	0.559	Q1	BMJ Publishing Group

) regarding research on AI in education conducted using qualitative methods (2014–2024) reveals broad dissemination across high-impact publications in the fields of education, technology, and health. PLOS One leads in total citations (337,945), while Computers and Education: Artificial Intelligence stands out for having the highest impact factor (SJR 3.227; Q1 of 2023). Journals such as the Journal of Medical Internet Research, Education and Information Technologies, and BMC Medical Education also play a key role in the dissemination of knowledge within this domain. The presence of major academic publishers such as Springer Nature, Elsevier, and MDPI underscores the consolidation of this research area within globally relevant scientific forums, encompassing both highly indexed journals and specialized publications in education and educational technology.

The analysis of the most highly cited documents (Table 3) reveals a growing concern with the integration of AI in education through the use of qualitative methods. The literature is characterized by a diverse and multidisciplinary scientific output, with systematic reviews emerging as the most influential contributions. Notably, article [78], with 1268 citations, stands out as one of the most impactful reviews on AI in education. This work, alongside others, such as [79] on AI education policy and [80] on teacher interactions with ChatGPT, highlights a trend toward understanding AI as a pedagogical complement rather than a replacement for traditional teaching. The presence of qualitative studies and systematic reviews published in high-impact journals reflects a strong interest in evaluating both the benefits and limitations of AI technologies in varied educational contexts.

Of the 630 qualitative studies analyzed, a clear thematic concentration was observed around generative AI applications. Overall, 31.8% of the works focus on tools such as ChatGPT, chatbots, Generative Artificial Intelligence (GenAI), image generation, or generative adversarial networks, demonstrating a notable predominance of this subcategory compared to other AI-based educational technologies. In comparison, studies on intelligent tutoring systems represent only 0.38%, while those addressing AI-enhanced augmented reality and virtual reality represent 0.96% and 1.34%, respectively. Research on personalized learning solutions (5.17%) and automated assessment systems (3.26%) are also comparatively underrepresented.

These data show that recent qualitative research has primarily focused on generative applications, especially ChatGPT, possibly due to their rapid adoption and social visibility in educational contexts. It should also be noted that more than half of the studies (52.1%) were not initially tagged under any of the predefined categories. An analysis of the most frequent terms in their titles, abstracts, and keywords reveals that they are also largely aligned with the central themes of the analysis: education, AI addressed in a general way (attitudes, opinions, perspectives, etc.), learning, data analysis, and research.

These results suggest that this technology occupies a central position in the current research agenda. The limited presence of applications other than GenAI suggests that the findings related to the experience of teachers and students may not be generalizable to all forms of AI in education, but rather specific to those tools—such as ChatGPT—that have achieved a greater degree of implementation and visibility.

Furthermore, while the topic is not monolithic, research on AI in education intersects with other disciplines such as data science, psychology, and higher education. For instance, the review on Moodle in [81] underscores the role of virtual learning environments in AI implementation, while [82] analyzes tweets related to education during the pandemic and emphasizes the value of text mining in understanding perceptions of AI. This disciplinary convergence suggests that the integration of AI in education is shaped by a multitude of factors—including technological infrastructure, education policy, and stakeholders’ perceptions—thus reinforcing the relevance of qualitative approaches to explore these complex dynamics in depth.

Table 3. The 12 most relevant documents in research on AI in education using qualitative methods in the years 2014–2024.

	Title	Author(s)	Source	Year	Citations
1	Artificial Intelligence in Education: A Review	Chen, L., Chen, P., and Lin, Z. [78]	IEEE Access	2020	1268
2	The Diversity-Innovation Paradox in Science	Hofstra, B., Kulkarni, V. V., Sebastian Munoz-Najar. S., He, B., Jurafsky, D., and McFarland, D. A. [83]	Proceedings of the National Academy of Sciences of the United States of America	2020	614
3	A comprehensive AI policy education framework for university teaching and learning	Chan, C.K.Y. [79]	International Journal of Educational Technology in Higher Education	2023	400
4	Large language models in education: A focus on the complementary relationship between human teachers and ChatGPT	Jeon, J. and Lee, S. [80]	Education and Information Technologies	2023	242
5	A comprehensive review on deep learning-based methods for video anomaly detection	Nayak, R., Pati, U. C., and Das, S. K. [84]	Image and Vision Computing	2021	196
6	Examining thematic similarity, difference, and membership in three online mental health communities from reddit: A text mining and visualization approach	Park, A., Conway, M., and Chen, A. T. [85]	Computers in Human Behavior	2018	161
7	Sentiment analysis and topic modeling on tweets about online education during COVID-19	Mujahid, M., Lee, E., Rustam, F., Washington, P. B., Ullah, S., Reshi, A. A., and Ashraf, I. [82]	Applied Sciences	2021	158
8	Creation and Evaluation of a Pretertiary Artificial Intelligence (AI) Curriculum	Chiu, T. K. F., Meng, H., Chai, C. S., King, I., Wong, S., and Yam, Y. [86]	IEEE Transactions on Education	2022	155
9	A systematic review on trends in using Moodle for teaching and learning	Gamage, S. H. P. W., Ayres, J. R., and Behrend, M. B. [81]	International Journal of STEM Education	2022	140
10	The use of ChatGPT in the digital era: Perspectives on chatbot implementation	Limna, P., Kraiwanit, T., Jangjarat, K., Klayklung, P., and Chocksathaporn, P. [87]	Journal of Applied Learning and Teaching	2023	130
11	Evaluating performance of biomedical image retrieval systems-An overview of the medical image retrieval task at ImageCLEF 2004–2013	Kalpathy-Cramer, J., de Herrera, A. G. S., Demner-Fushman, D., Antani, S., Bedrick, S., and Müller, H. [88]	Computerized Medical Imaging and Graphics	2015	126
12	Socio-technical imaginary of the fourth industrial revolution and its implications for vocational education and training: a literature review	Avis, J. [89]	Journal of Vocational Education and Training	2018	94

Table 3. The 12 most relevant documents in research on AI in education using qualitative methods in the years 2014–2024.

	Title	Author(s)	Source	Year	Citations
1	Artificial Intelligence in Education: A Review	Chen, L., Chen, P., and Lin, Z. [78]	IEEE Access	2020	1268
2	The Diversity-Innovation Paradox in Science	Hofstra, B., Kulkarni, V. V., Sebastian Munoz-Najar. S., He, B., Jurafsky, D., and McFarland, D. A. [83]	Proceedings of the National Academy of Sciences of the United States of America	2020	614
3	A comprehensive AI policy education framework for university teaching and learning	Chan, C.K.Y. [79]	International Journal of Educational Technology in Higher Education	2023	400
4	Large language models in education: A focus on the complementary relationship between human teachers and ChatGPT	Jeon, J. and Lee, S. [80]	Education and Information Technologies	2023	242
5	A comprehensive review on deep learning-based methods for video anomaly detection	Nayak, R., Pati, U. C., and Das, S. K. [84]	Image and Vision Computing	2021	196
6	Examining thematic similarity, difference, and membership in three online mental health communities from reddit: A text mining and visualization approach	Park, A., Conway, M., and Chen, A. T. [85]	Computers in Human Behavior	2018	161
7	Sentiment analysis and topic modeling on tweets about online education during COVID-19	Mujahid, M., Lee, E., Rustam, F., Washington, P. B., Ullah, S., Reshi, A. A., and Ashraf, I. [82]	Applied Sciences	2021	158
8	Creation and Evaluation of a Pretertiary Artificial Intelligence (AI) Curriculum	Chiu, T. K. F., Meng, H., Chai, C. S., King, I., Wong, S., and Yam, Y. [86]	IEEE Transactions on Education	2022	155
9	A systematic review on trends in using Moodle for teaching and learning	Gamage, S. H. P. W., Ayres, J. R., and Behrend, M. B. [81]	International Journal of STEM Education	2022	140
10	The use of ChatGPT in the digital era: Perspectives on chatbot implementation	Limna, P., Kraiwanit, T., Jangjarat, K., Klayklung, P., and Chocksathaporn, P. [87]	Journal of Applied Learning and Teaching	2023	130
11	Evaluating performance of biomedical image retrieval systems-An overview of the medical image retrieval task at ImageCLEF 2004–2013	Kalpathy-Cramer, J., de Herrera, A. G. S., Demner-Fushman, D., Antani, S., Bedrick, S., and Müller, H. [88]	Computerized Medical Imaging and Graphics	2015	126
12	Socio-technical imaginary of the fourth industrial revolution and its implications for vocational education and training: a literature review	Avis, J. [89]	Journal of Vocational Education and Training	2018	94

4.3. Leadership and International Collaboration in the Study of AI in Education with Qualitative Methods (RQ3)

Figure 5 shows that the output of scientific research articles on AI in education using qualitative methods has increased significantly in several countries from 2014 to 2022. Countries such as the United States, China, and the United Kingdom stand out for their high levels of production. The United States has maintained steady growth, while China has shown a more pronounced increase in recent years, surpassing other countries in 2021 and 2022. The United Kingdom has also increased its production, albeit at a more moderate pace. These data confirm that these countries are leading research in this field, reflecting greater investment and focus on developing AI in education using qualitative methods.

The network shown in Figure 6 represents international collaboration in research on AI in education supported by qualitative methodologies. In this visualization, node size indicates the scientific output of each country, while the links between nodes reflect the strength of co-authorship relationships. The United States (145 documents, 2435 citations, link strength 70) and the United Kingdom (58 documents, 733 citations, link strength 56) emerge as the principal research hubs, serving as highly connected nodes with multiple countries. China is also a key contributor (63 documents, 1725 citations, link strength 29), although its level of international connectivity is lower compared to that of the U.S. and the U.K.

Regional collaboration patterns are evident, including a strong interconnection among European countries (Germany, Spain, Italy, and Switzerland), an Asian block led by China and Hong Kong, and an Anglo-speaking alliance encompassing the U.S., the U.K., Australia, and Canada. Countries with lower research output, such as Bangladesh, Ghana, and the Philippines, appear on the periphery of the network, reflecting more limited involvement in international collaborations. The distribution of the network underscores the concentration of research efforts in countries with greater academic and technological resources, highlighting the need for strategies aimed at fostering the inclusion of emerging nations in the global AI-in-education research agenda, particularly in studies using qualitative methods.

The historical direct citation network (Figure 7) illustrates the relationships between key articles on AI in education that employ qualitative methods and have been instrumental in shaping the field. Articles with the highest Local Citation Score (LCS) and Global Citation Score (GCS) have exerted significant mutual influence and have played a central role in guiding subsequent research, reflecting both the evolution and continuity of the domain. One notable example is reference [90], published in 2021 in the Journal of Science Education and Technology, which explores how augmented observation supports multimodal representational thinking by applying deep learning to decode students’ complex constructions. With an LCS of 1 and a GCS of 26, this work has been cited by studies focusing on the use of AI to enhance science education.

Similarly, [91] is an article published in 2022 in the Journal of Research in Science Teaching investigating the application of machine learning for the automated evaluation of scientific models, emphasizing the use of deep neural networks and natural language processing. With an LCS of 1 and a GCS of 60, this paper has influenced further research on inclusive assessment and scientific modeling. Published in 2021, [92] is an article in IEEE Transactions on Learning Technologies that analyzes large collections of open-ended MOOC feedback using Latent Dirichlet Allocation (LDA) topic modeling and qualitative analysis. With an LCS of 5 and a GCS of 49, this work has been foundational in advancing automated text analysis for improving distance education.

Additionally, from 2022, [93] is an article published in the Journal of Information Technology Education Research, in which the authors conducted a systematic review of the use of AI in English language teaching and identified a growing use of AI platforms such as chatbots, voice recognition systems, machine translation engines, and intelligent tutoring systems in learning English as a foreign language. With an LCS of 2 and a GCS of 56, this study has informed research on the challenges and impacts of AI in language education. Lastly, the 2023 article [94], published in Digital Health, evaluates the need for chatbot-based instructional programs for nursing students learning patient history-taking. With an LCS of 3 and a GCS of 7, this study has contributed to the growing body of literature on AI use in health education.

4.4. Keywords and Thematic Trends in Qualitative Research on AI in Education (RQ4)

The keyword co-occurrence network in research on AI in education using qualitative methods (Figure 8) reveals several meaningful relationships among key terms, which can be organized into thematic clusters. These clusters span a wide range of topics, from the evaluation of educational technologies and clinical practices to curriculum development and higher education, integrating advanced AI techniques with qualitative analysis to address diverse issues in both educational and healthcare contexts.

Cluster 1 (red) includes terms such as “machine learning” (105 occurrences, link strength 716), “qualitative analysis” (96 occurrences, link strength 852), “systematic review” (21 occurrences, link strength 143), and “technology” (16 occurrences, link strength 102). This suggests a focus on leveraging advanced AI and qualitative methods to assess and review educational technologies, particularly in health and education settings. Cluster 2 (green) groups terms such as “artificial intelligence” (283 occurrences, link strength 1669), “chatgpt” (83 occurrences, link strength 434), “deep learning” (50 occurrences, link strength 311), and “higher education” (44 occurrences, link strength 108). This cluster emphasizes the integration of advanced AI and language models in higher education, with particular attention being paid to student motivation and language learning.

Cluster 3 (blue) comprises terms like “clinical practice” (21 occurrences, link strength 259), “interview” (34 occurrences, link strength 447), “thematic analysis” (47 occurrences, link strength 523), and “questionnaire” (48 occurrences, link strength 610). These terms are primarily associated with clinical and health-related studies, where interviews and thematic analysis are used to explore the experiences and practices of patients and professionals in medical environments. Cluster 4 (yellow) includes terms such as “curriculum” (32 occurrences, link strength 352), “medical education” (77 occurrences, link strength 849), “focus groups” (15 occurrences, link strength 204), and “students” (65 occurrences, link strength 319). This cluster reflects research on medical education and curriculum development, where focus groups and surveys are commonly employed to evaluate educational programs and student perceptions.

Finally, Cluster 5 (purple) centers on “human experiment” (28 occurrences, link strength 322), indicating a smaller set of studies involving experimental designs with human participants, suggesting a niche but emerging research area within the broader field.

Analysis of the keywords reveals a clear focus of articles on higher and professional education levels, especially in university, medical, and nursing contexts. This trend suggests that academic production focuses on specialized training and the preparation of professionals in technical and health areas. In contrast, the absence of explicit mentions of primary, secondary, or non-university vocational training indicates scant attention is being paid to these levels in the analyzed literature. This omission could reflect a gap in educational research using qualitative methods, where initial and technical levels, fundamental to comprehensive educational development, do not receive the same visibility or priority in current studies.

The thematic map of AI in education using qualitative methods (Figure 9) is divided into four quadrants, each representing different levels of topic relevance and development. These quadrants help illustrate how topics are distributed and evolve within this research field. In the upper-left quadrant are niche topics, which are highly developed but of lower general relevance. A prominent example is creativity, which, while specialized, suggests targeted applications where creative processes are essential, such as the design of interactive and personalized educational materials, along with specific types of machine learning such as contrastive learning, adversarial machine learning, and federated learning.

The upper-right quadrant contains motor themes—topics that are both highly developed and relevant—indicating that they are central and dynamic in the current research landscape. This includes areas related to healthcare professionals, particularly the use of semi-structured interviews and patient care. These themes reflect the meaningful application of AI to enhance healthcare services and qualitative inquiry in medical contexts. Semi-structured interviews are essential for capturing detailed insights into the effectiveness of AI in such environments, while patient care benefits from AI-driven educational tools that support informed decision-making by professionals.

The lower-left quadrant houses emerging or declining topics, those with low relevance and development. These may be in the early stages of exploration or losing traction. They include text analysis, natural language processing, teachers, and higher education. Despite being less developed, these emerging areas offer promising avenues for improving education through AI, such as developing automated tutoring systems and analyzing student responses.

Finally, the lower-right quadrant features basic themes, which are topics that are highly relevant but less developed. These represent foundational areas with significant potential for future research. They include AI, qualitative research, machine learning, and human factors. AI and machine learning are core techniques for developing intelligent educational systems that adapt to students’ individual needs, while qualitative research provides rich insights into user experiences and perceptions. The inclusion of human factors highlights the importance of addressing human interaction and ethical considerations in the integration of AI in educational contexts.

5. Discussion

This study provides a detailed and up-to-date overview of the development of scientific research output related to the use of AI in education through qualitative methodologies. First, regarding the chronological evolution over the past decade (RQ1), scientific output indexed in Scopus has shown sustained growth, with a notable increase beginning in 2018 and peaking in 2021 and 2023. This trend aligns with the accelerated pace of technological advancement and a growing interest in adapting AI to educational contexts, especially following the COVID-19 pandemic, which catalyzed the digital transformation of educational systems [23]. The consolidation of AI in education has sparked interest in exploring its implications through qualitative lenses, which offer deeper, context-rich insights into experiences, perceptions, and educational processes [95].

In terms of influential authors and sources (RQ2), journals such as Computers and Education: Artificial Intelligence, Education and Information Technologies, and BMC Medical Education have emerged as leading platforms at the intersection of education, technology, and methodological innovation. Scholars such as Di Zou, Alejandra J. Magana, and Jonathan Kantor have made high-impact contributions, addressing issues such as curriculum design, automated feedback, and chatbot integration. The most prolific institutions are located in the United States, the United Kingdom, China, and Germany (RQ3), reflecting the concentration of technological resources and academic capital in these regions. This hegemony is grounded in robust infrastructures, access to research funding, and strong global networks. Their leadership is further strengthened by active participation in international collaborations, which foster complex, intercultural, and geographically diverse qualitative research. Similar patterns are observed in other fields, such as medicine [96], psychology [97], and architecture, where globally connected institutions lead high-quality knowledge production. International collaboration is thus a key driver of methodological innovation, cross-validation of findings, and inclusive perspectives [98].

Thematic and keyword analyses in the domain of qualitative AI research in education (RQ4) reveal an expanding and diversifying field, with a strong focus on AI’s potential to support data analysis, personalize learning, and enhance educational and professional practices. There is a marked orientation toward higher education and clinical settings, where AI is valued for its ability to support curriculum development, assistive training, and evidence-based decision-making, as highlighted by [4]. The prominence of participatory methodologies and qualitative interviews reinforces their role as central strategies in exploring technology-mediated educational experiences. While machine learning and human factors are recognized as essential, they remain underdeveloped, indicating substantial room for growth and future consolidation [7]. The field is characterized by a coexistence of established, emerging, and developing topics, reflecting a dynamic and evolving research landscape.

Healthcare-related studies stand out as driving forces, particularly those involving semi-structured interviews and patient care, which demonstrate AI’s positive impact on decision-making and professional training [96]. Simultaneously, topics such as creativity, federated learning, and personalized content design represent specialized lines with strong potential in specific educational contexts. Despite the high relevance of machine learning and human-centered factors, deeper exploration is needed, pointing to a promising future research agenda that integrates technological innovation with pedagogical understanding and ethical sensitivity. This thematic interconnection signals a growing trend toward more personalized and humanized AI-mediated learning environments.

6. Conclusions

Our SLR reveals a significant and sustained evolution in the scholarly exploration of AI within education through qualitative approaches. This trend reflects a broader transformation in how educational technologies are not only adopted but also critically examined in relation to pedagogical, social, and ethical dimensions. The growing body of research indicates an increasing commitment within the academic community to interpret and contextualize the impact of AI beyond its technical functionalities.

Leadership in this domain is concentrated among prominent authors and institutions that have become central nodes in the production and dissemination of knowledge, signaling the emergence of a specialized, globally interconnected research community. Their contributions reflect a maturing field that is moving toward a deeper engagement with the complex realities of AI-mediated education. The presence of high-impact publications across disciplines, particularly in technology-enhanced learning, health education, and the social sciences, demonstrates the interdisciplinary relevance of this research area. Leading academic publishers and journals serve as critical platforms, not only for sharing findings, but also for shaping the epistemic foundations of qualitative inquiry into AI in education. This consolidation of scholarly output and visibility underscores the central role of qualitative research in informing responsible and reflective integration of AI technologies in educational systems worldwide.

Although this study offers a broad and detailed overview of the field, it relies exclusively on the Scopus database. While other platforms, such as Web of Science, ERIC, or Google Scholar, could provide additional insights, Scopus was selected due to its extensive coverage in education and social sciences and its strong integration with bibliometric analysis tools like Bibliometrix. Initial comparisons revealed substantial overlap between Scopus and Web of Science, which reduced the added value of including both in this exploratory phase. Nonetheless, this decision may limit the comprehensiveness of the corpus. Future research should consider incorporating multiple databases and complementary search strategies to enable more systematic and in-depth analyses. Additionally, as the present study relies on quantitative bibliometric metrics, it does not assess the methodological quality or social impact of the selected publications. Another limitation of the study lies in the search strategy employed, which was based on general terms such as “artificial intelligence” and “education.” This approach may have excluded relevant studies that use more specific terminology associated with AI subcategories, such as intelligent tutoring systems, natural language processing, or generative AI. Future research should consider refining the keywords, guided by more precise thematic classifications, to achieve a more comprehensive retrieval of the existing literature. To address this, future lines of inquiry should include qualitative content analysis and thematic exploration of representative documents.

In addition, in the future, it is essential to delve deeper into emerging areas such as creativity, human factors, AI ethics, and personalized learning using qualitative approaches that allow for exploration of the experiences, perceptions, and tensions experienced by educational stakeholders in their interaction with these technologies. Likewise, the need to promote empirical studies with multicultural and multilingual perspectives is highlighted, highlighting the diversity of contexts, languages, and forms of knowledge from which AI is adopted and redefined in different educational settings. This research should aim not only to broaden the geographical and cultural representation in scientific research output but also to question dominant epistemic frameworks, fostering the development of educational AI that is more inclusive, situated, and sensitive to local realities.